The present invention relates to an information processing system arranged to have plural processing units like computers connected with each other, and in particular to a method of sending or receiving information among the processing units through a network and the apparatus thereof.
Conventionally, an information processing system for transferring information through a network such as a distributed processing system has employed one of two systems, that is, a Unicast system and a Broadcast system. The Unicast system is arranged so that a processing unit for transmitting data specifies an address of a processing unit for receiving the data, thus implementing one-to-one data transfer. On the other hand, the Broadcast system is arranged so that data sent from a processing unit connected in the network is received by all of the other processing units connected in the network.
To send data as specifying plural processing units contained in a network, for example, a communication protocol referred to as Internet Protocol includes the method of dividing a node address into a field indicating a subnetwork and a field indicating a host. In addition to the Unicast and the Broadcast, sending of data to all the nodes contained in the subnetwork, that is, grouping of destinations (Multicast communication) is allowed to be implemented. The method for creating or releasing a group in the Multicast communication at any time is disclosed in J-P-A-4-157940. In this method, a node requesting participation in a group sends a request of participation with a group address as a parameter to another node through the Broadcast communication. A right of representation for the group is possessed by the node having first requested participation in the group. The node operates to process a request of participation sent from another node and manage the nodes in the group. In this system, any node is able to participate in a group at any time or belong to plural groups.
If the Unicast or the Multicast is used as means for communicating with another processing unit in the network, for providing communication, it is possible to establish a connection between the subject processing units. The connection termed herein includes a physical connection between the processing units as well as an overall communicating protocol for supporting transmission authentication to re-transmitting of a message if any error takes place in the signal between the subject processing units. In this case, a response for authenticating the receipt of a message is given back from the receiving processing unit to the sender processing unit. This response makes great contribution to enhancing reliability of data transmission. The Broadcast communication, on the other hand, requires that a connection be established among all of the processing units. This makes the load that is burdened on the network very excessive. Hence, ordinarily, the communication is done without having to establish the connection. It means that the Broadcast communication has lower reliability of data transmission than the Unicast or the Multicast.
In the aforementioned techniques, any one of the communicating methods is selected for transmitting data or any one of the processing units (or the groups of the processing units in the case of the Multicast) is determined as a destination by a user program (or application program) running on the processing unit. To transmit data on the Unicast or the Multicast, the user program is required to grasp an arrangement of the system on the user program. The change or expansion of the system arrangement cannot be easily done. To solve this shortcoming, the U.S. Pat. No. 4,366,479, for example, discloses the method for broadcasting data from the transmitting side with a description defined according to the content of the data called the content code and entrusting judgement as to data requirement to the receiving processing unit. This method, however, provides lower reliability of data transmission than the Unicast or the Multicast, because all of the data are broadcast.
The J-P-A 4-157940, on the other hand, discloses a technique of enabling grouping or regrouping of the processing units running on a network. If, however, a new node (processing unit) is added to the network, the existing nodes cannot recognize information such as an address of the newly added node. That is, for the communication method requiring that a destination address be specified, like the Unicast or the Multicast, online addition or deletion of a node is not allowed.
The conventional Multicast has the capability of restricting a sending range of data by defining a destination address when registering a group. The receiving node receives all of the data sent to the group(s) to which it belongs. In actuality, however, the processing units contained in the group for the Multicast do not need all of the data sent to the group. That is, the conventional Multicast communication is considered as a system for broadcasting data in a group and does not have exceptional sending efficiency in the group.